Skip to main content

SC ClickUp Ticket Audit Playbook

Purpose​

This is the decision layer behind two slash commands: /audit-client-tickets (one client, deep) and /audit-stale-tickets (whole workspace, staleness-focused — see Workspace-wide mode). It defines how open ClickUp tickets are classified into stale (safe to retire), needs assignee response (live work that went quiet), needs-update (status/field/tag fix), or OK — and what disposition to recommend for each.

It exists so the audit's judgment is consistent, cited, and improving:

  • Cited, not invented. Every check below traces to an existing SC standard or working rule. The audit applies those rules; it does not make up new ones.
  • Improving. The command reads this playbook at the start of every run and appends learnings to the Run Log at the end. Thresholds here are defaults to be tuned from what the audit gets wrong in practice.

How the command uses this doc

  1. Load this playbook + follow its pointers to the source standards.
  2. Classify each open ticket using Staleness signals and Needs-update checks.
  3. Recommend a disposition via the Disposition decision tree (reversible-first).
  4. After the run, append a dated Run Log entry and bump this doc's version if any heuristic/threshold changed.

Scope + guardrails​

  • One client per run for /audit-client-tickets. That command refuses multi-client / whole-workspace audits. The workspace sweep is a separate, narrower command — see Workspace-wide mode.
  • Open tickets only are classified; completed tickets are surfaced only as cold-storage archive candidates (see below).
  • Read-only until approval. No ClickUp mutation happens before Rohit approves the report.
  • Reversible-first. Stale tickets are retired by cancel + archive (recoverable), never hard-deleted except for obvious junk with explicit per-ticket approval.
  • no-edit-existing-tickets boundary. "Needs update" means metadata, status, tags, and close-with-rationale only. A ticket whose description or scope is wrong may be edited in place only while it is to do with no time logged; once work has started, recommend supersede with a new ticket instead ([[feedback_close_with_rationale_pattern]]).

Source standards (the audit cites these)​

AreaSource
Status model, invariants, terminal statesSC Task Data Standard §1, §3
Terminal-status handling + >30-day cold-storage sweepTask Management Workspace Setup SOP §3.1, §5.6
10 field-completeness linter rulesContractor Task Quality — Pre-Work Standard
AC walk-through + quality tagging at reviewContractor Task Quality — Review Standard
One assignee per ticket/subtasksingle-assignee-per-ticket
Client field / client/<slug> tag disciplineclient-identifier-discipline
Zero open decisions in an execution ticketexecution-vs-decision-tickets
Tag must be in taxonomySC ClickUp Tag Taxonomy, clickup-tag-applied-or-blocked
Thin-ticket shape, no filesystem paths, canonical linksMarkdown Task File → ClickUp SOP §3–§4
Due dates in ESTcanonical-timezone-for-due-dates-est
Assignee update cadence + dwell ceilingClickUp Ways of Working §11

Last assignee touch​

S1, S4, and S7 measure time since the assignee last touched the ticket — not date_updated.

date_updated bumps on any write, including description edits made by the Creator, an audit run, or an automation. A ticket can read "updated yesterday" while the person who owns it has not looked at it in two months. Measuring it produces false negatives on exactly the tickets that need flagging.

Last assignee touch = the most recent of:

  • a comment authored by the current assignee,
  • a status change made by the assignee,
  • a time entry logged by the assignee.

Comments authored by anyone else — the Creator, a reviewer, an audit — do not reset the clock. They are input to the assignee, not evidence the assignee acted.

Practically, the audit reads clickup_get_task_comments and filters on user.id == assignee.id. Where comment history is empty, time-in-current-status from clickup_get_bulk_tasks_time_in_status is the fallback measure.

Staleness signals​

A ticket is a stale candidate if any signal fires. Thresholds are defaults — tune via the Run Log.

#SignalDefault triggerRecommended disposition
S1Dormant + not startedLast assignee touch older than 30 days AND status still to do AND zero time trackedCancel + archive
S2Empty placeholderNo/near-empty description AND no assignee AND no due dateCancel + archive (or hard-delete if truly junk, per-ticket approval)
S3Superseded / duplicateA comment, superseded_by, or a near-identical sibling shows the work moved elsewhereCancel + archive with a supersession comment linking the replacement
S4Abandoned in progressin progress or blocked with last assignee touch older than 45 daysFlag for owner decision (reopen intent vs cancel) — do not auto-cancel
S5Completed, coldstatus complete AND date_closed/date_updated older than 30 daysCold-storage archive (per Setup SOP §5.6) — not "stale", just tidy-up
S6Meta-ticket fulfilledTicket's deliverable is other tickets (imperative-to-create-tickets title) and those tickets now existClose (complete if children were created from it; cancelled if superseded) — it is done, not dormant
S7Assignee silenceOpen ticket with exactly one assignee, status to do / in progress / blocked, and last assignee touch older than 7 calendar daysNEEDS ASSIGNEE RESPONSE — not cancel. Silence signals neglect, not obsolescence. Nudge, then escalate per Ways of Working §11.
S8Orphaned assignmentAssignee user ID is absent from clickup_get_workspace_members (deactivated or removed member)Reassign to a current member, or park (no assignee + no due date). The ticket looks owned in every view and is not.
S9Declared done, never closedAn assignee comment states the work is finished, but status is still to do / in progressVerify against the Evidence bullet, then close as complete. Confirms the "event-satisfied closes" pattern first logged 2026-08-03.

Anti-false-positive guards:

  • A ticket created within the threshold window is never stale, regardless of status.
  • A ticket the assignee touched within the window is never stale. A comment from anyone else does not clear the flag — see Last assignee touch.
  • cancelled tickets are already terminal — leave visible (they're a signal), do not re-flag.
  • A parked/deferred ticket with an explicit future due date or a blocked-on:* tag is not dormant — it's waiting. Route to needs-update only if the block is untagged.
  • A ticket explicitly paused in a comment — naming why, when it resumes, and what unblocks it — is not S7. This is the correct handling of deferred work and must not be penalized.
  • S7 never fires on a parked ticket: parked means no assignee, and S7 requires exactly one.

Composite staleness score​

A workspace sweep returns hundreds of tickets. Rank them so the report is a queue, not a wall. Score each flagged ticket:

ComponentPoints
Days of assignee silencemin(days / 7, 8)
Days in current status beyond its ceiling (Ways of Working §11)min(excess_days / 10, 6)
Past due+3, plus min(days_overdue / 14, 4)
No due date on an assigned, non-parked ticket+2
Status is in progress (actively misrepresents state)+2
S8 orphaned assignment+5
S9 declared done+6 (cheapest possible win — closing costs nothing)

Report in descending score. The score orders work; it never decides disposition — the decision tree does that.

Workspace-wide mode​

/audit-stale-tickets sweeps every space instead of one client list. The differences from a client audit:

  • Staleness-only. It runs S1, S4, S7, S8, S9 and the U-checks that need no client context (U1, U2, U3, U4, U7). It skips U5/U6 (client field/tag correctness) — those need per-client knowledge and belong to /audit-client-tickets.
  • Never applies in bulk. A client audit can apply an approved batch. The workspace sweep applies per-ticket only, on explicit per-ticket approval, because a wrong call is repeated across every client at once.
  • Cohort reporting. Tickets bulk-created together (same day, same source meeting, near-identical dwell) are reported as one row with a count, not N rows. A 14-ticket cohort is one decision. This is the single biggest determinant of whether the report is usable.
  • Unassigned tickets are reported separately, not scored. Per the parked-backlog convention, no-assignee + no-due-date is deliberate. The sweep counts them and flags only those that look unintentionally parked (high/urgent priority, or referenced by live work).
  • Pagination is mandatory. clickup_filter_tasks caps at 100 per page; Client Delivery alone exceeds that. Loop on has_more / next_page until exhausted, and state the scanned count in the report — a silently truncated sweep reads as "everything is fine."

Needs-update checks​

Metadata/status/tag problems on an otherwise-valid ticket. Each maps to a source rule.

Ticket-not-ready tagging. U4 (wrong or inferred assignee, not a parked ticket), U8, U9, U10 and U13, and S3 when the duplicate was created rather than inherited, are defects in how the ticket was written. Alongside the fix, recommend qa:ticket-not-ready with a comment naming the check, e.g. qa:ticket-not-ready — U13. The tag is a recommendation like every other audit output and is applied only on approval.

#CheckFires whenFix (metadata only)
U1Status drift — time vs TODONon-zero time tracked but status still to doAdvance to in progress (Task Data Standard invariant)
U2Blocked without reasonStatus blocked but no blocked-on:* tagAdd the correct blocked-on:* tag (or unblock)
U3Complete with open QAcomplete while a qa-review-* / am-review-* subtask is still openReopen parent to in progress until the review subtask closes
U4Assignee count ≠ 1Zero or multiple assignees, unless the ticket is intentionally parked (see below)Set exactly one assignee (split work into siblings if shared) — [[single-assignee-per-ticket]]

Parked-backlog convention (adopted by Rohit 2026-08-06): a ticket that is deliberately deprioritized is left with no assignee and no due date — that combination is the parked-backlog signal, not a U4/U7 defect. Audits must not re-assign or re-date parked tickets; they flag a parked ticket only if it looks unintentionally parked (e.g., high/urgent priority, or referenced as active by a live workstream). Reactivation = assign one owner + set a due date. This replaces the earlier "speculative due dates on client-gated tickets" churn (2026-08-03 run) — the blocked-on:client tag proposal was superseded by this simpler mechanism. | U5 | Missing Client field | Client Delivery ticket with no Client dropdown value | Set Client to the client — [[client-identifier-discipline]] | | U6 | Tag drift | A tag not in the taxonomy, or client/<slug> tag ≠ Client field | Remove/correct; if a needed tag is unprovisioned, surface as blocking (don't silently skip) | | U7 | Missing scheduling fields | No due date and/or no priority on an active execution ticket | Set due (EST) + priority | | U8 | Missing effort / Evidence | No estimate; or no Evidence: bullet on a client-facing deliverable | Advisory flag (Evidence blocking for client-facing, advisory internal) | | U9 | Open decision in execution ticket | Body contains a "decide X" (explicit, or implicit — an execution step presupposing an undefined artifact) that gates the work | Recommend a separate decision ticket + sequence. Rewrite in place only if the ticket is to do with no time logged | | U10 | Evergreen / vague title | Open-ended "Maintain X"-style title with no discrete deliverable | Recommend reframe as a discrete action or recurring checklist (edit in place if to do with no time logged, otherwise supersede); if the underlying milestone already happened, route to stale instead |

| U13 | Unverified claim stated as fact | The body asserts current system state (a setting, a routing destination, a count, a configuration) with no verification date and no trace to a source | Verify it live — the audit already holds MCP access. If true, add a dated confirmation comment. If false, the ticket is sending the assignee at a phantom: supersede it |

U13 is the one check the audit is uniquely placed to run. It reads live tickets with platform access, so it can test a claim the writer only assumed. See #unverified-claims-rule — that rule is Blocking but has no machine check, and this is its enforcement surface.

U9 and any description/scope defect are supersede-not-edit once work has started. The audit reports them; it rewrites a body only on a to do ticket with no time logged, and only with approval.

Disposition decision tree​

For each open ticket, in order:

  1. S9 (declared done) → CLOSE AS COMPLETE after verifying the Evidence bullet. Check this first: a finished ticket should not be audited for staleness or polished for metadata.
  2. S8 (orphaned assignment) → REASSIGN OR PARK. Check this second: every downstream signal measures the assignee, and there isn't one.
  3. Any needs-update check fires but the ticket is clearly live work → NEEDS UPDATE (list the specific fixes).
  4. Any staleness signal S1–S4 fires and no guard applies → STALE → cancel + archive (hard-delete only S2-junk, per-ticket approval).
  5. Both stale and fixable? Prefer STALE — don't polish a ticket that should be retired.
  6. S7 alone (assignee silence, but the ticket is otherwise well-formed and still wanted) → NEEDS ASSIGNEE RESPONSE. This bucket is a nudge list, not a retirement list; it never auto-cancels.
  7. complete + cold (S5) → ARCHIVE candidate (separate bucket, low-risk tidy-up).
  8. Otherwise → OK, leave.

S7 vs S1/S4. S7 fires at 7 days, S1/S4 at 30/45. A ticket crossing 7 days is quiet; a ticket crossing 30 is probably unwanted. Reporting them in the same bucket destroys that distinction — quiet tickets get cancelled and real backlog gets nudged. Keep the buckets separate.

Ambiguous (S3/S4, or "is this still wanted?") → do not auto-act; present to Rohit with the evidence and let him decide.

Report shape​

  • Drive (source of truth): Client_Delivery/<slug>/work-in-progress/YYYY-MM/YYYY-MM-ref-<slug>-clickup-ticket-audit.md, grouped by disposition; per row: ticket link, title, assignee, status, age / last-updated, signal(s) fired, recommended action, one-line rationale.
  • Inline (chat): a terse triage table of only the decision-requiring rows ([[feedback_terse_status_reports]]) — not a full re-dump of OK tickets.

Apply phase (only after approval)​

  • Field/status/tag fixes per the needs-update list.
  • Stale → set status cancelled + a close-with-rationale comment: why closing / when to reopen / where the protection now lives; then archive.
  • Cold completes → archive.
  • Hard-delete → only on explicit per-ticket "yes".
  • Never rewrite an existing description/scope.

Run Log​

Append one entry per run (newest last). This is the self-improvement record — capture what the audit got wrong so the next run is better.

### YYYY-MM-DD — <client-slug>
- Open tickets scanned: N (stale: a · needs-update: b · archive-cold: c · OK: d)
- Rohit overrides: <what he reclassified and why>
- New stale patterns / false positives: <observation>
- Threshold/heuristic changes made this run: <e.g. S1 30d→45d for this client, or none>

2026-07-02 — rotary-sports (first run, test bed)​

  • Open tickets scanned: 29 (stale: 0 · needs-update: 10 + minor priority gap on 15 fresh · likely-done→close: 1 · cold-complete archive candidates: ~17 · OK: 17).
  • Report: Client_Delivery/rotary-sports/work-in-progress/2026-07/2026-07-ref-rotary-sports-clickup-ticket-audit.md.
  • Rohit overrides: (pending — recommendations not yet applied).
  • New patterns / candidate heuristics for the next version:
    • Meta-ticket-fulfilled. A ticket whose deliverable is other tickets (e.g. 86ajbbff5 "Create ClickUp tickets for follow up items") is done once its children exist — treat as close, not stale. Detect by imperative-to-create-tickets titles + a source-doc pointer. → candidate signal S6.
    • Evergreen/vague title. "Maintain X" / open-ended titles (e.g. 86aj1upt2 "Maintain Rotary go live plan") aren't discrete execution tickets — flag to reframe as a discrete action or a recurring checklist. → candidate needs-update check U10 (ties to execution-vs-decision-tickets discreteness).
    • Lapsed-due ≠ stale. Many tickets here are dormant + past-due but still clearly wanted; a past due date is a needs-update (refresh due) signal, not a staleness signal on its own. S1 correctly requires dormancy and to do and zero time — keep that AND, don't loosen to "past due → stale."
    • Bulk-create priority gap. Freshly bulk-created subtasks commonly ship without a priority (U7) even when otherwise well-formed. Worth a create-time checklist item rather than an audit finding — noted for the ticket-creation flow, not a threshold change here.
  • Threshold/heuristic changes made this run: none to S1–S5/U1–U9 (defaults held). Added S6/U10 as candidates — promote to the tables after a second client confirms they generalise.

2026-07-03 — rotary-sports (follow-up learning)​

  • Not a full audit run — logged from a ticket rework driven by the 2026-07-02 Tanu–Rohit division-of-labor reset.
  • Assignee-correctness ≠ has-an-assignee. The 2026-07-02 audit marked ~15 Rohit-assigned tickets "OK" because they had a single assignee — but the team model is "Tanu executes, Rohit reviews", so most should have been Tanu's. → candidate needs-update check U11: assignee vs. division-of-labor — an audit should read the client's current ownership model (from the digest / latest internal sync) and flag tickets assigned to the wrong role, not just missing assignees. Requires an ownership-model input the audit doesn't currently take — note for the command's next iteration.
  • Decision hidden inside an execution ticket. "Send Aman the pricing doc" assumed a pricing model existed; it didn't (not in contract). U9 (open decision in execution ticket) should also catch implicit decisions — an execution step that presupposes an undefined artifact. Reinforces U9; no table change.

2026-07-27 — rotary-sports (second full run)​

  • Open tickets scanned: 57 (cancelled: 16 · closed complete: 2 · needs-update applied: 31 · OK: 8). Post-apply active count: 39.
  • Report: Client_Delivery/rotary-sports/work-in-progress/2026-07/2026-07-ref-rotary-sports-clickup-ticket-audit-2.md (suffix -2; the 7/02 report keeps its name).
  • Rohit overrides / calls: cancel the Canva-postcard rebuild (superseded by persisted render sources); keep tune-up-page and LC-trial tickets with refreshed metadata; cancel (not hard-delete) the three empty placeholders.
  • Confirmed patterns:
    • S6 (meta-ticket fulfilled) and U10 (evergreen title) both fired again and survived Rohit's review — promoted from candidates to the main tables this run (v1.3). U11 (assignee vs division-of-labor) remains a candidate: it needs an ownership-model input the command doesn't take yet.
    • Meeting-generated ticket batches age into supersession. Two bulk batches (6/30 call notes 86ajbb*, 7/09 call notes 86ajg23*) were later replaced wholesale by structured workstreams; ~10 of the 16 cancels came from those batches. Heuristic: unassigned, no-priority tickets created the day of a client call, whose topic matches a later structured parent, are S3 by default.
    • Salvage before cancel. One junk ticket carried the only copy of the client's FB/IG handle links — copied to the surviving ticket before cancelling. Check cancel candidates for unique data worth salvaging.
  • Threshold/heuristic changes made this run: S6 + U10 promoted to the tables; no numeric threshold changes (S1 30d / S4 45d held).

2026-08-03 — rotary-sports (third run, light review)​

  • Open tickets scanned: 39 (closed complete: 4 — event-satisfied · needs-update applied: 16 due-date fixes · stale/cancel: 0 · OK: 19). Note: 16 cancelled tickets from 7/27 still appear in the open filter — cancelled is a done-type but not closed-type status in this space; harmless but inflates raw counts.
  • Rohit overrides: none (plan approved as recommended).
  • Dominant churn insight — speculative due dates on client-gated tickets. ~20 dues were refreshed on 7/27; ~14 of those lapsed again by 8/3 because the gating action sits with the client (FB access, markup edits, payment, trial feedback), not SC. Refreshing dates is administrative churn that no amount of auditing fixes.
  • Recommended practice change (pending Rohit's adoption): client-gated tickets get status blocked + a blocked-on:client tag instead of a guessed due date; due dates only on SC-controlled work; the weekly client call agenda reviews the blocked set. Prereqs: blocked-on:client must be added to the tag taxonomy (PR) and provisioned in the Client Delivery space (known gap — it failed as unprovisioned in prior sessions). If adopted, U7 changes semantics ("missing due" fires only on SC-controlled tickets) → version bump at that point.
  • New patterns: event-satisfied closes — 4 tickets were completed by the flow of client work (calls, supersessions) without anyone closing them. The audit is currently the only mechanism that reaps these; the weekly-call blocked-set review would catch them sooner.

2026-08-11 — whole workspace (first workspace-wide run)​

  • Scope: all 10 spaces, all assignees, assigned + open only. Scanned 100 assigned open tickets — Tanu 49, Rohit 48, orphaned 3. Parveen / Ravish / the second Rohit record hold zero.
  • ~59 of 100 have not moved status in over 30 days.
  • Reports: Shared drives/sc-docs-storage/2026-08-ref-clickup-stale-sweep.md + companion 2026-08-ref-tanu-clickup-ticket-review.md. Nothing applied — awaiting Rohit's per-ticket approval.
  • Tooling gotcha (important, cost a wrong count). clickup_filter_tasks silently drops results when passed more than one assignee ID. ["94074999","81938975"] returned 36 Client Delivery tickets; ["94074999"] alone returned 37 (it lost 86ajqhhwv). Always query one assignee per call and reconcile against an unfiltered space pull. The first draft of the Tanu review shipped "48 open" on the multi-ID result and had to be corrected to 49.
  • S7 validated. Fired correctly on 86ahe17nr (assignee comment 2026-06-17, in progress 75d). Correctly did not fire on 86ajw9hyq (explicitly paused with a comment naming why/when/what-unblocks) or on the Banyan Tree tickets created within the window. The explicit-pause guard is load-bearing — without it the one correctly-managed ticket in the set would have been flagged.
  • S8 validated. 3 tickets in upscale-legal > active-work assigned to 100012322 "Ravish Kumar Singh" and 100012325 "Zakir Hussain", neither in get_workspace_members. Note 100012322 is a duplicate of current member 82112671 "Ravish Kumar" — S8 catches stale duplicate accounts, not just departures. Dwell 315–320d.
  • S9 validated. 86aj71wjy — assignee commented "Testing is done. Email sent." 2026-07-10; still to do 47 days later. Confirms the 2026-08-03 "event-satisfied closes" observation and justifies promoting it to a signal.
  • New pattern — QA-review subtasks as a silent bottleneck. Three qa-review-* subtasks open 39 days, all overdue, against the 2-business-day reviewer SLA (QA Reviewer SOP §4.3). Because Invariant 3 blocks parent closure on an open QA subtask, each one holds a parent open regardless of whether the work is done. Candidate needs-update check U12: QA-review subtask past reviewer SLA — distinct from U3 (complete-with-open-QA), which catches the opposite error. Promote after a second run confirms.
  • New pattern — the reviewer is also the biggest assignee. 97 of 100 assigned tickets belong to two people, one of whom is the default QA Reviewer. Relates to the still-uncommitted U11 (assignee vs. division-of-labor) and to the absent WIP threshold.
  • Confirmed, third time: meeting-generated batches dominate the stale set (moots-law go-live 6 × 102d, KraneWorks call items 8 × 108d, KraneWorks follow-ups 4 × 50–55d). This is now the largest single source of stale tickets across three runs. The fix is upstream of the audit — meeting action items need the pre-work bar applied at creation.
  • Parked-backlog convention was only ever applied to rotary-sports. ~15 tickets elsewhere would qualify. Recommend applying workspace-wide.
  • Threshold changes this run: none. S7 (7d) / S8 / S9 are new this version and held on first contact.

2026-08-24 — moots-law (first run, pre-meeting prep)​

  • Open tickets scanned: 11 (close-as-complete: 3 · cancel/delete-candidates: 3 · keep-discuss: 4 · keep-internal: 1 · needs-update-only: 0). ~18 cold completes (S5) noted as archive batch.
  • Report: Client_Delivery/moots-law/work-in-progress/2026-08/2026-08-ref-moots-law-clickup-ticket-audit.md. Applied 2026-08-24 same-day on Rohit's approval: 3 closed complete + 3 cancelled with rationale comments; no hard deletes (no per-ticket approval sought — cancel is reversible); keep-set metadata refresh deferred to post-meeting. Post-apply active count: 5.
  • Rohit overrides: none (applied as recommended).
  • New patterns / observations:
    • Zero comments across all 11 open tickets — last-assignee-touch had no comment signal at all; date_updated/date_created was the only measure. On lists where nobody comments, S9 evidence must come from outside ClickUp (sibling completed tickets, email threads, uptime monitors) — all three close-as-complete calls here were externally evidenced.
    • Vendor-exit supersession (S3 variant): two tickets died because the vendor they operate on (SearchAtlas) was exited company-wide. Client-list audits should check active vendor-drawdown state before classifying platform-specific tickets — the supersession note lives in a roadmap doc, not in ClickUp.
    • Same-workstream ticket pair (two hero tickets from different sessions, 3 weeks apart) — recommend supersede-both-with-one rather than picking a survivor, since each holds part of the context.
  • Threshold/heuristic changes made this run: none (S1 30d / S4 45d held; S9 fired 3x on external evidence).

2026-08-25 — rotary-sports (fourth run, pre-meeting prep)​

  • Open tickets scanned: 31 (raw filter returned 52 — 21 lingering cancelled rows, the known done-but-not-closed quirk). Close-as-complete: 1 · cancel: 2 · keep-discuss: 8 · keep-internal: 8 · parked (left): 11 · needs-update-only: 1 (C3 blocked without tag — blocked-on:client still unprovisioned).
  • Report: Client_Delivery/rotary-sports/work-in-progress/2026-08/2026-08-ref-rotary-sports-clickup-ticket-audit.md. Applied same-day on Rohit's approval: 86ajt9zgt closed complete + 86ajc00rp/86ajc00rw cancelled with rationale comments. Post-apply active count: 28. Keep-set due-date refresh deferred to post-meeting (real commitments only — the 8/3 speculative-date churn lesson held: all 11 dates applied 8/3 had lapsed again).
  • Rohit overrides: none (applied as recommended).
  • New patterns / observations:
    • Live-platform UI as S9 evidence. Extends the moots zero-comment lesson: the one close-complete was evidenced by the GHL Integrations page (Instagram shows connected), not by anything in ClickUp. Pre-meeting audits should budget a short live-platform pass — it also caught two digest §8 items (475-line retirement, default-number switch) done-but-still-marked-OPEN in canon. Canon lag cuts both ways: verify claimed-open the same way we verify claimed-exists.
    • Register-governed creative rounds supersede per-version qa-review tickets (S3 variant). Once a piece has a revision register with a round sign-off table, the standing qa-review-<piece> ticket is redundant — the register row is the gate. Expect this wherever the Client Creative Revision SOP is in force.
  • Threshold/heuristic changes made this run: none (S1 30d / S4 45d held).

2026-08-27 — kraneworks (first run, pre-meeting prep; applied 2026-08-28)​

  • Open tickets scanned: 19 (raw filter returned 30 — 11 lingering cancelled, the known quirk). Close-as-complete: 4 · cancel: 2 · ambiguous-close: 2 (left open for Rohit) · keep-discuss: 5 · keep-internal: 5 · parked: 1 · needs-update fixes on keeps: due set on 1 (86aht7dt4 → meeting date), due cleared on 1 (86ahvd9ea → parked properly).
  • Report: Client_Delivery/kraneworks/work-in-progress/2026-08/2026-08-ref-kraneworks-clickup-ticket-audit.md. Applied 2026-08-28 on Rohit's approval — 4 completed + 2 cancelled with rationale comments. Post-apply active count: 13.
  • Rohit overrides: none (applied as recommended; ambiguous pair deferred to him).
  • New patterns / observations:
    • Zero comments on 13 of 19 — confirms the moots-law lesson at a second client: on low-comment lists, S9 evidence is external. All 4 closes here were evidenced by live site HTML, sub-account UI, Drive artifacts, or email, none by ClickUp signals.
    • Live-platform pass caught a paid resource leak, not just ticket state: an IVR test phone number still provisioned on the client's live sub-account, billing monthly while the deciding ticket (86aj71wkx) sat client-gated. Pre-meeting audits should check for provisioned-but-undecided paid resources (numbers, seats, add-ons) attached to open decision tickets — the ticket costs nothing, the resource does.
    • Meeting-batch supersession confirmed, fourth run (86ajg23t3/86ajg23ru from the 7/09 call batch).
    • Client-gated cluster (5 tickets) mapped 1:1 onto the meeting agenda — the audit's keep-discuss bucket doubling as the agenda skeleton worked well for the third straight prep-pack.
  • Threshold/heuristic changes made this run: none (S1 30d / S4 45d held; no version bump).

Version history​

DateVersionChangeAuthor
2026-07-021.0Initial playbook — staleness signals, needs-update checks, disposition tree, report shape, apply-phase rules, Run Log. Backs the /audit-client-tickets command.Symphony Core Systems Team
2026-07-021.1First run (rotary-sports) logged. Added candidate signals S6 (meta-ticket-fulfilled) + U10 (evergreen/vague title) from run learnings — pending a second client before promotion to the tables.Symphony Core Systems Team
2026-07-031.2Follow-up learning from the 2026-07-02 division-of-labor rework: candidate check U11 (assignee vs. division-of-labor) + reinforced U9 to catch implicit decisions.Symphony Core Systems Team
2026-07-271.3Second rotary-sports run applied (57 scanned; 16 cancelled, 2 completed, 31 updated). Promoted S6 (meta-ticket fulfilled) and U10 (evergreen title) from candidates to the main tables — both re-fired and survived owner review; promotion approved by Rohit with the run. U9 wording extended to name implicit decisions. Added meeting-batch-supersession and salvage-before-cancel guidance to the Run Log.Symphony Core Systems Team
2026-08-061.4Adopted the parked-backlog convention (no assignee + no due date = deliberately parked; U4/U7 exempt). Supersedes the 2026-08-03 blocked-on:client tag proposal. First application: 13 rotary-sports tickets (phone/IVR + SMS clusters, AI add-on, C4) parked per Rohit's direction.Symphony Core Systems Team
2026-08-111.5Added the Last assignee touch measure and retuned S1/S4 to use it instead of date_updated (which bumps on Creator/audit writes and hides real dormancy). New signals: S7 assignee silence (7d, new NEEDS ASSIGNEE RESPONSE bucket), S8 orphaned assignment (assignee absent from workspace members), S9 declared done, never closed — S9 promotes the "event-satisfied closes" pattern logged 2026-08-03. Added a composite staleness score for ranking large sweeps and a Workspace-wide mode section backing the new /audit-stale-tickets command. Disposition tree reordered so S9 and S8 resolve before staleness. Driven by the 2026-08-11 review of assigned tickets, which found 6 assignee comments across 48 open tickets and 3 genuinely-blocked tickets carrying no blocked status.Symphony Core Systems Team
2026-09-071.4Added needs-update check U13 (unverified claim stated as fact) and named the audit as the enforcement surface for the new #unverified-claims-rule in the pre-work standard, which is Blocking but has no machine check. Driven by a KraneWorks drafting session where six tickets carried three premises nobody had checked; every form rule would have passed. U11 and U12 remain candidates.Symphony Core Systems Team
2026-10-051.6The supersede-not-edit boundary is gated on status: a to do ticket with no time logged may have its description or scope fixed in place; supersession applies once work has started. Follows the same change to no-edit-existing-tickets.Symphony Core Systems Team
2026-10-221.7Checks U4, U8, U9, U10, U13 and S3 now recommend the new qa:ticket-not-ready tag alongside their fix, so authoring defects found by the audit are counted against the CreatorSymphony Core Systems Team